The Japanese sample audio for “Zero-Shot In-Context Generation” does not match the context.

[1m-N00b]

Describe the bug
The Japanese sample audio for “Zero-Shot In-Context Generation” does not match the context.
Everything up to 14 seconds of audio is incoherent Japanese and does not exist in the context; after 14 seconds, the content is consistent with the context.
Similar Issues:#303

To Reproduce
Steps to reproduce the behavior:

1. Go to Zero-shot In-context Generation.
2. Select the sample in the lower right corner as shown in the screenshot.

Expected behavior
Audio files should be placed after 14 seconds or the context should be added correctly.
(However, there seems to be a problem on the audio generation side.)

Screenshots

Desktop (please complete the following information):

• Browser: firefox,chromium

[aluminumbox]

well this should be modified if we have more Japanese data, or try change Chinese character to Japanese character